The occurrence of intraspecific predation in Pacific hake diet appears to be highly variable, in contrast to previous estimates that cannibalism forms a significant portion of the Pacific hake diet - up to 30% (CITE). Stomachs gathered from the hake acoustic trawl survey, however, do not show a high rate of intraspecific predation, with only 16 instances of cannibalism out of 3995 stomachs collected from 2005-2019: a rate of 0.4%.
Across the CCTD, hake depredation comprises less than 30% of their diet, with the proportion of prey hake in predator hake stomachs reaching over 75% in the 1990s (Fig. 1 & 2). In the dataset, Pacific hake is the largest contributor to their diet by weight (Sup. Fig. 1). For the entire timeseries, hake depredation is on the youngest hake, with the vast majority on age-1 and younger and with no records of predation on hake older than 5 years (Fig. 2).
Figure 1: Individual hake stomachs collected from 1980-2020, containing Pacific hake (in green) or only other food items (in purple).
Figure 2: Dirichlet-weighted diet proportion by weight of Pacific hake intraspecific predation by predator hake age and prey hake age for three periods: all years in the dataset, 1991-1999, and the acoustic trawl survey from 2005-2019.
The timing and location of cannibalism is also variable, with potential evidence in some years of a higher occurrence of cannibalism at the southern end of the sampling area (below approximately 42°N) and later in the year. This trend is visible in 1998 and 1999 and without the geographic bias in 1995 and 1997 (Fig. 3).
Figure 3: Hake stomachs containing Pacific hake (in green) or only other food items (in purple) by location per year and by sampling month per year.
Without diet data (operating as a single-species model), the CEATTLE model matches well with the Stock Synthesis 3.0 (SS3) estimates for spawning stock biomass (SSB), total biomass, and recruitment for the timeseries (Fig. 5). When the mean rates of intraspecific predation (i.e. the rates for all years as in Fig. 2) are included, there is not an appreciable change in SSB, with a slight overall decrease in the estimate. For total biomass, the model run with cannibalism deviates more widely from both the single-species CEATTLE model and the SS3 runs. For recruitment, the number of individuals is much higher for the CEATTLE run with cannibalism as the model “creates” recruits to satisfy the higher rates of predation on year 1 hake (Fig 5.)
Figure 4: Spawning stock biomass (mt), total biomass (mt), and recruitment (n) for Pacific hake from CEATTLE run in single-species mode with no predation, with the estimate of mean cannibalism, and the Stock Synthesis 3.0 run from the 2020 stock assessment.
The predation pressure on the youngest hake in the model is incorporated through predation mortality (M2), which is added to the residual mortality (M1) and fishing mortality to determine the total mortality coefficient used to calculate numbers at age per year. M1 is estimated as age- and sex-invariant. When the model is run with intraspecific predation, the predation mortality on the age 1 hake increases dramatically, but there is little effect on the mortality of any other age group (Fig. 6).
Figure 5: Residual (M1) and predation (M2) mortality by age and year for CEATTLE run with the estimate of mean cannibalism.
The result of this excess mortality of the youngest hake is also apparent in the numbers-at-age, with a large increase in the number of age-1 hake estimated for the CEATTLE run with cannibalism, compared to the single-species model and the SS3 estimates (Fig. 6). The numbers-at-age are also highly variable across the timeseries (Fig. 6), with the largest increases in age-1 hake in the CEATTLE model with cannibalism occurring in 1981 and 1985, corresponding to the peaks visible in the estimates for total biomass, number of recruits, and mortality (Fig. 5 & 6).
Figure 6: Mean and standard deviation of numbers-at-age from CEATTLE run in single-secies mode with no predation, with the estimate of mean cannibalism, and the Stock Synthesis 3.0 run from the 2020 stock assessment.
The model was also fit with the Dirichlet-reweighted diet proportions for the two periods with hake data - 1988-1999 and 2005-2019 (Fig. 1 & 2). The estimates for SSB, total biomass, and recruitment were significantly higher when the model was run for only 1988-1999 with the mean diet proportion for that period and the estimates were slightly higher for the same over 2005-2019 (Fig. 7). The estimate of natural mortality follows the same trend, with mortality doubled for the run with only 1988-1999 included and a much smaller relative increase in mortality detected for 2005-2019 (Fig. 8). Across all model runs there are similar trends in all population dynamics visible, though on varying scales.
Figure 7: Spawning stock biomass (mt), total biomass (mt), and recruitment (n) for Pacific hake from CEATTLE run with the estimate of mean cannibalism for the entire period, in single-species mode with no predation, for 1988-1999 with the estimate of mean cannibalism for that period, and for 2005-2019 with the estimate of mean cannibalism for that period.
Figure 8: Residual (M1) and predation (M2) mortality by age and year for CEATTLE run with the estimate of mean cannibalism for the entire period, for 1988-1999 with the estimate of mean cannibalism for that period, and for 2005-2019 with the estimate of mean cannibalism for that period.
The overall mean proportion of intraspecific predation in the Pacific hake diet potentially belies the impact of cannibalism on hake population dynamics. To explore the effects of differing levels of cannibalism in CEATTLE on population dynamics, the model was fit with a diet proportion of between 0.5 and 80% prey hake by weight (representing the extreme end of the potential cannibalism as seen in the 1990s), with all prey consisting of year 1 hake. Generally, with increasing intraspecific predation, the amount of individuals “created” by the model increases, with the estimates for biomass and recruitment for the stomach contents data relatively similar to an overall rate of 10% intraspecific predation by weight (Fig. 7).
Figure 9: Runs of CEATTLE with varying levels of intraspecific predation (0.5% to 80%) and the mean cannibalism by weight from the Pacific hake stomach contents datasets.
The amount of predation mortality also follows the expected trend, with increasing amounts of excess mortality produced with the proportion of hake predation (Fig. 9). This effect is especially evident in the 1980s, corresponding with the highest peaks in biomass and recruitment for the higher rates of cannibalism as well (Fig. 8 & 9). It is important to note that for both the model runs with the empirical data and for the sensitivity testing, the diet proportion remained constant across years.
Figure 10: Residual (M1) and predation (M2) mortality by age an year for CEATTLE run with varying levels of inrraspecies predation (0.5% to 80%).
While previous studies of Pacific hake diets have recorded a high proportion of intraspecific predation and have noted the role of cannibalism in population dynamics, applying cannibalism to a model of hake population dynamics does not appear to have a significant effect on current population dynamics. There is significant uncertainty introduced, however, by the highly variable rates of intraspecific predation.
While the different sampling protocols and varying temporal and geographic coverage make it difficult to compare the rates of intraspecific predation across the timeseries, it does appear that more cannibalism was detected in the 1990s and potentially futher south at those times. Given the large sample size of stomachs taken during the Pacific hake acoustic-trawl survey conducted from 2005-2019 (Fig. 1), it is reasonably clear that, at least in the summer months currently, hake cannibalism has not played a significant role in the diet of hake in the California Current in the last 15 years. As hake are a general predator, shifts in their diet can be expected (CITE). However, given variable sampling methods and sample sizes earlier in the timeseries, it is unclear whether the higher rates of hake predation recorded at the time are representitive of a period of greater hake depredation or result from sampling bias.
While further analysis of the trends in hake diets was outside the scope of this paper, there is evidence of a similar trend in the proportion of hake present in the diets of California sea lions, with higher rates of hake depredation detected in the 1980s and 1990s than in the 2000s onwards (Fig. 9). The sea lion diet dataset consists of scat records from San Clemente and San Nicholas Islands in Southern California and has much more even coverage for the timeseries. The sea lion data were also collected quarterly, with some evidence that more hake were recorded in their diet in the Spring and Fall, especially in the earlier years (Fig. 9). These data can maybe serve as corroboration of some change in Pacific hake population dynamics manifesting in their depredation by both sea lion and their conspecifics.
Figure 11: Island-wide hake anomaly in California sea lion scat collected on San Clemente Island and San Nichols Island in the Spring, Summer, Fall, and Winter from 1980-2021. Provided by Dr. Alexandra Curtis.
Because CEATTLE is fit to the fishery and survey catch data, the model adapts to the excess mortality created by cannibalism by creating additional age-1 hake. This suggests that with some amount of cannibalism likely occurring in the population, a single-species stock assessment of Pacific hake likely underestimates the number of young hake (ages 0-5) in the population, especially ages 0-1. However, due to the highly variable recruitment and the fishery starting at age 3 (CITE), this overestimation likely does not have a large effect on the exploited biomass, especially if the low levels of recent cannibalism are assumed.
Any risk to the population from intraspecific predation would result from a combination between diet pressure and lower recruitment due to other factors. Natural mortality does increase with cannibalism, as seen across the model runs, including the run with the low, mean level of intraspecific predation. Natural mortality in the stock assessment is assumed to be 0.23 for the youngest hake (age 0 in the assessment, age 1 in CEATTLE), then 0.21 for subsequent ages (CITE). All model runs except for the sensitivity test with 0.5% cannibalism predict higher rates of natural mortality for the youngest hake (Fig. 5, 8, 10).
Assuming a higher rate of natural mortality than is predicted in the stock assessment, it could be expected that intraspecific predation would play a role hake’s susceptibility to and recovery from a low recruitment event. To simulate the effects of multiple drivers, CEATTLE could be used to simulate the population at depressed recruitment levels. Additionally, to simulate the potential signal of changes in cannibalism along with population dynamics for stock assessment and the fishery, a management strategy evaluation (MSE) could also be conducted with CEATTLE (CITE, Grant in prep).